GxA results version 4

1 Results version history

• GxA results version 1 @ 2017-09-19
• GxA results version 0

2 Relevant reports

• Report 23: Results on inference of ancestry

3 Description of results

Outcomes:

• 7 continious traits
  • 1 binary trait was not considered, as mixed models take much more time

Ancestry data:

• We disgarded the local ancestry information computed by Margaret in 2012 (Hapmap II ref. panel, hg19)
• We re-computed the ancestry
  • another reference panel of 1,000 Genome haplotypes (v5, hg19)
  • the same LAMD-LD tool and settings
• Results:
  • the number of segments increased 153,046 from to 229,384
  • no articfacts in local ancestry along the genome
  • sanity check: indentical estimations of global ancestries for the older and newer computations

Association analysis

• We tested all 229,384 ancestry segments for association
• We fited linear mixed models (LMM) by using Gaston (not lme4qtl)
• Outcome/trait values were imputed (the sample size is remained 3,300)

Multi-trait association analysis

• 7 traits: FEV1pp_utah, FEV1_utah, FVCpp_utah, FVC_utah, FEV1_FVC_utah, log_pctEmph_Slicer, TLCpp_race_adjusted
• threshold: 3
• In additional runs with threshold to 2.5
  • the ancestry signal at Chr 10 (effective BF) becomes singificant (passes effective BF)
  • the diagonal entries of covariance matrix are slightly more far away from 1

3.1 Significannt thresholds

3.1.1 Bonferroni threshold

0.05 / 229384

[1] 2.179751e-07

4 Results

4.1 Interaction with SmokCigNow

4.1.1 Manhattan plots

4.1.2 Chr 11, PARVA gene

4.1.2.1 Fine-mapping

Table of top SNP and other SNPs from the same ancestry segment (chr11acs76906):

snp	zscore	pval	chr	pos	acs	num_snps	size	trait
rs933920	2.928	0.003413	11	12481110	chr11acs76906	21	36273	FEV1pp_utah
rs4756884	-2.557	0.01055	11	12474098	chr11acs76906	21	36273	FEV1pp_utah
rs4756866	-2.369	0.01781	11	12448846	chr11acs76906	21	36273	FEV1pp_utah
rs10831815	2.076	0.03792	11	12446924	chr11acs76906	21	36273	FEV1pp_utah
rs10741590	2.056	0.03976	11	12463318	chr11acs76906	21	36273	FEV1pp_utah
rs1344615	-1.901	0.05725	11	12454046	chr11acs76906	21	36273	FEV1pp_utah
rs11022358	1.647	0.09951	11	12470449	chr11acs76906	21	36273	FEV1pp_utah
rs10831827	-1.603	0.1089	11	12483197	chr11acs76906	21	36273	FEV1pp_utah
rs1365403	1.571	0.1161	11	12476231	chr11acs76906	21	36273	FEV1pp_utah
rs1346220	1.543	0.1229	11	12480087	chr11acs76906	21	36273	FEV1pp_utah

NCBI quiery (the latest dbSNP build and latest reference genome):

Query	Chromosome	Marker	Class	Gene	Alleles	Major	Minor	MAF	BP	AncestralAllele
rs933920	11	rs933920	snp	PARVA	A/G	A	G	0.0377	12459563	A,A,A,A,A,A
rs4756884	11	rs4756884	snp	PARVA	A/G	A	G	0.3768	12452551	A,A,A,A,A
rs4756866	11	rs4756866	snp	PARVA	A/G	A	G	0.031	12427299	A,A,A,A,A,A
rs10831815	11	rs10831815	snp	PARVA	A/C	C	A	0.4018	12425377	C,C,C,C,C,C
rs10741590	11	rs10741590	snp	PARVA	C/T	C	T	0.1791	12441771	C,C,C,C,C,C
rs1344615	11	rs1344615	snp	PARVA	C/T	C	T	0.1214	12432499	C,C,C,C,C,C
rs11022358	11	rs11022358	snp	PARVA	C/T	T	C	0.2612	12448902	C,C,C,C,C,C
rs10831827	11	rs10831827	snp	PARVA	A/G	G	A	0.2586	12461650	G,G,G,G,G,G
rs1365403	11	rs1365403	snp	PARVA	G/T	G	T	0.0859	12454684	T,T,T,T,T,T
rs1346220	11	rs1346220	snp	PARVA	C/T	C	T	0.0617	12458540	T,T,T,T,T,T

4.2 Interaction with SmokCigNow0_15

4.2.1 Manhattan plots

4.2.2 Chr 11

4.2.2.1 Fine-mapping

• ANKRD49 gene: A four-gene signature from NCI-60 cell line for survival prediction in non-small cell lung cancer
  • Differential analysis in case-control data based on microarray data
• GPR83 gene: Genetic Associations with Obstructive Sleep Apnea Traits in Hispanic/Latino Americans
  • Obstructive sleep apnea (OSA) is the most common type of sleep apnea and is caused by complete or partial obstructions of the upper airway. It is characterized by repetitive episodes of shallow or paused breathing during sleep, despite the effort to breathe, and is usually associated with a reduction in blood oxygen saturation. These episodes of decreased breathing, called “apneas” (literally, “without breath”), typically last 20 to 40 seconds. wikipedia

4.2.2.1.1 COPD and Obstructive Sleep Apnea

COPD; Mesh term

All MeSH Categories
  Diseases Category
  Respiratory Tract Diseases
    Lung Diseases
      Lung Diseases, Obstructive
        Pulmonary Disease, Chronic Obstructive

Obstructive Sleep Apnea; Mesh Term

Categorization 1:

All MeSH Categories
  Diseases Category
    Respiratory Tract Diseases
      Respiration Disorders
        Apnea
          Sleep Apnea Syndromes
            Sleep Apnea, Obstructive

Categorization 2:

All MeSH Categories
  Diseases Category
    Nervous System Diseases
      Sleep Wake Disorders
        Dyssomnias
          Sleep Disorders, Intrinsic
            Sleep Apnea Syndromes
              Sleep Apnea, Obstructive

4.2.3 Chr 10

4.2.3.1 Fine-mapping

• BICC1 gene: This gene encodes an RNA-binding protein that is active in regulating gene expression by modulating protein translation during embryonic development. Mouse studies identified the corresponding protein to be under strict control during cell differentiation and to be a maternally provided gene product.

4.2.4 Chr 2

4.2.4.1 Fine-mapping

5 QQ plots

5.1 Interaction with SmokCigNow

5.2 Interaction with SmokCigNow0_15

6 Quality control on multi-trait analysis

• 7 traits: FEV1pp_utah, FEV1_utah, FVCpp_utah, FVC_utah, FEV1_FVC_utah, log_pctEmph_Slicer, TLCpp_race_adjusted
• threshold: 3

6.1 Interaction with SmokCigNow

6.2 Interaction with SmokCigNow0_15

7 Models formulas

trait	ver	formula
FEV1pp_utah	9	imp_FEV1pp_utah ~ Age_Enroll + gender + Duration_Smoking + log_CigPerDaySmokAvg + SmokCigarNow + SmokCigNow + SmokCigNow0_15 + acs_global2 + (0 + dummy(SmokCigNow0_15Rev)|rsid1) + (0 + dummy(SmokCigNowRev)|rsid2) + (1|ccenter)
FEV1_utah	9	imp_FEV1_utah ~ Age_Enroll + Age_Enroll2 + gender + Height_CM + Duration_Smoking + log_CigPerDaySmokAvg + SmokCigarNow + SmokCigNow + SmokCigNow0_15 + acs_global2 + (0 + dummy(SmokCigNow0_15Rev)|rsid1) + (0 + dummy(SmokCigNowRev)|rsid2) + (1|ccenter)
FVCpp_utah	9	imp_FVCpp_utah ~ Age_Enroll + gender + Duration_Smoking + log_CigPerDaySmokAvg + SmokCigarNow + SmokCigNow + SmokCigNow0_15 + acs_global2 + (0 + dummy(SmokCigNow0_15Rev)|rsid1) + (0 + dummy(SmokCigNowRev)|rsid2) + (1|ccenter)
FVC_utah	9	imp_FVC_utah ~ Age_Enroll + Age_Enroll2 + gender + Height_CM + Duration_Smoking + log_CigPerDaySmokAvg + SmokCigarNow + SmokCigNow + SmokCigNow0_15 + acs_global2 + (0 + dummy(SmokCigNow0_15Rev)|rsid1) + (0 + dummy(SmokCigNowRev)|rsid2) + (1|ccenter)
FEV1_FVC_utah	9	imp_FEV1_FVC_utah ~ Age_Enroll + Age_Enroll2 + gender + Duration_Smoking + log_CigPerDaySmokAvg + SmokCigNow + SmokCigNow0_15 + acs_global2 + (0 + dummy(SmokCigNow0_15Rev)|rsid1) + (0 + dummy(SmokCigNowRev)|rsid2) + (1|ccenter)
log_pctEmph_Slicer	9	imp_log_pctEmph_Slicer ~ Age_Enroll + gender + Duration_Smoking + log_CigPerDaySmokAvg + SmokCigarNow + SmokCigNow + SmokCigNow0_15 + acs_global2 + (0 + dummy(SmokCigNow0_15Rev)|rsid1) + (0 + dummy(SmokCigNowRev)|rsid2) + (1|ccenter) + (1|scanner_gr)
TLCpp_race_adjusted	9	imp_TLCpp_race_adjusted ~ Age_Enroll + gender + Duration_Smoking + log_CigPerDaySmokAvg + SmokCigarNow + SmokCigNow + SmokCigNow0_15 + acs_global2 + (0 + dummy(SmokCigNow0_15Rev)|rsid1) + (0 + dummy(SmokCigNowRev)|rsid2) + (1|ccenter)

8 R code for the Omnibus test

Test	Statistics	Distribution	Reference
Omnibus	\(z^T \Sigma^{-1} z\)	\(\chi^2(p)\)	(Aschard et al., in preparation)
sumZ	\((1^T z)^2 / 1^T \Sigma^{-1} 1\)	\(\chi^2(1)\)	(Province et al, 2013)

  jtab <- switch(method,
    "omnibus" = {
      P1 <- tcrossprod(Z, Sigma_inv)
      P2 <- P1 * Z # element-wise multiplication
      z2score <- rowSums(P2)
      
      data_frame(z2score = z2score) %>%
        mutate(pval = pchisq(z2score, df = ncol(Z), lower = FALSE))
    },
    "metacp" = {
      P1 <- tcrossprod(Z, Sigma_inv)
      
      z2score <- (rowSums(P1))^2 / sum(Sigma_inv)
      
      data_frame(z2score = z2score) %>%
        mutate(pval = pchisq(z2score, df = 1, lower = FALSE))
    },
    "sum2" = {
      z2score <- (rowSums(Z))^2 / sum(Sigma)
      
      data_frame(z2score = z2score) %>%
        mutate(pval = pchisq(z2score, df = 1, lower = FALSE))
    },    
    stop("switch"))